A River of Bones: Wildebeest Skeletons Leave a Legacy of Mass Mortality in the Mara River, Kenya
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
ORIGINAL RESEARCH
published: 27 February 2020
doi: 10.3389/fevo.2020.00031
A River of Bones: Wildebeest
Skeletons Leave a Legacy of Mass
Mortality in the Mara River, Kenya
Amanda L. Subalusky 1,2* , Christopher L. Dutton 1,2 , Emma J. Rosi 3 , Linda M. Puth 2 and
David M. Post 2
1
Department of Biology, University of Florida, Gainesville, FL, United States, 2 Department of Ecology and Evolutionary
Biology, Yale University, New Haven, CT, United States, 3 Cary Institute of Ecosystem Studies, Millbrook, NY, United States
Animal carcasses can provide important resources for a suite of consumers, and
bones may provide a largely overlooked component of this resource, as they contain
a large proportion of the phosphorus (P) in a carcass and they can persist for decades
to millennia. We synthesized several datasets from our research in the Mara River,
in which annual mass drownings of wildebeest (Connochaetes taurinus) contribute
Edited by:
2.2 × 105 kg of bones per year, to examine the ecological role that bone could
Gary A. Lamberti, play in this river ecosystem and to prioritize research questions on the role of bones
University of Notre Dame, in aquatic ecosystems in general. We measured bone stoichiometry and used in-
United States
stream litterbags to measure bone decomposition rate, both of which varied by bone
Reviewed by:
Michael J. Vanni, type. Decomposition occurs as a two-stage process, with 15% of the mass being
Miami University, United States relatively labile and decomposing in 80–120 days and the remaining recalcitrant portion
Peter B. McIntyre,
Cornell University, United States
decomposing over > 80 years, leading to an estimated standing stock of 5.1 × 106 kg
*Correspondence:
bones in the river. We used mesocosm experiments to measure leaching rates from
Amanda L. Subalusky bones. Leachate from fresh bones was an order of magnitude higher in inorganic
asubalusky@ufl.edu nitrogen (N) than P; however, aged bones from the river leached much more P than
Specialty section:
N, which stimulated primary production. Biofilms growing on bones had five times
This article was submitted to greater chlorophyll a and two times greater organic matter than those growing on rocks,
Population and Evolutionary
although algal composition was not significantly different between the two substrates.
Dynamics,
a section of the journal Biofilms growing on bones also differed from biofilms on rocks in carbon (C) and N
Frontiers in Ecology and Evolution stable isotope signature. Mixing models suggest that biofilms on bones account for
Received: 01 April 2019 19% of macroinvertebrate and 24% of fish tissues in the Mara River, even months
Accepted: 03 February 2020
Published: 27 February 2020
after carcasses were present. In combination, these findings suggest that bones may
Citation:
influence nutrient cycling, ecosystem function, and food webs in the Mara River,
Subalusky AL, Dutton CL, potentially on decadal time scales. Bones may also be important in other aquatic
Rosi EJ, Puth LM and Post DM (2020)
ecosystems, and mass extirpations of large land mammals may have led to a loss of
A River of Bones: Wildebeest
Skeletons Leave a Legacy of Mass this resource. Large animal bones may play a unique role in ecosystems via their slow
Mortality in the Mara River, Kenya. release of limiting nutrients.
Front. Ecol. Evol. 8:31.
doi: 10.3389/fevo.2020.00031 Keywords: aquatic ecosystem, bone, carcass, decomposition, production, river, skeleton, stoichiometry
Frontiers in Ecology and Evolution | www.frontiersin.org 1 February 2020 | Volume 8 | Article 31
Subalusky et al. Bone Decomposition Influences Aquatic Ecosystems
INTRODUCTION of collagen, non-collagenous proteins, and lipids. As a result of
this structure, bones have a much higher proportion of P than
Animals can have myriad effects on biogeochemistry, nutrient soft tissues, with N:P ratios of < 1:1 (Elser et al., 1996; Subalusky
cycling, and ecosystem function through both direct and indirect et al., 2017), and bones can persist in ecosystems for decadal time
effects on trophic processes and through transport processes scales (Smith and Baco, 2003; Wenger et al., 2019). Because the
(Bauer and Hoye, 2014; Schmitz et al., 2018). Animals tend to scaling of bone and body size in terrestrial vertebrates is non-
aggregate in time and space, which can lead to biogeochemical linear, larger-bodied vertebrates have a much larger proportion
hot spots and hot moments (McClain et al., 2003), and animals of their total body mass comprised of bone than small-bodied
can move across ecosystem boundaries, which can transport animals (Prange et al., 1979; Elser et al., 1996). Altogether, these
resource-rich subsidies against natural gradients or at an elevated studies suggest bones may provide a long-term, P-rich resource,
rate along natural gradients (Subalusky and Post, 2018). Live particularly when bones result from carcasses of large vertebrates,
animals can play important roles in driving these dynamics raising questions about the role they may play in nutrient cycling
through nutrient assimilation and excretion, during which and consumer dynamics.
animals can serve as sinks for some elements by assimilating The fate of bones in an ecosystem is largely influenced
them in their body tissue (Kitchell et al., 1979; Atkinson by environmental context. Bones in terrestrial ecosystems
et al., 2016; Nobre et al., 2019). After death, animal carcasses are subject to decomposition via exposure to sun and rain,
may play an important role as a nutrient source by liberating fungus, and foraging by animals that can consume bones,
limiting nutrients (Bump et al., 2009; Beasley et al., 2012; including rodents and larger animals such as hyenas. Bone
Keenan et al., 2018). persistence in tropical, terrestrial ecosystems is on the scale
Animal carcasses provide a complex and heterogeneous of several decades (Behrensmeyer, 1978; Trueman et al., 2004;
resource for an array of consumers. Animal carcasses can result Western and Behrensmeyer, 2009; Ross and Cunningham, 2011).
from annual or seasonal mortality associated with normal life Bone decomposition in temperate and arctic latitudes, which
history, selective drivers of mortality (e.g., disease, predation, are cooler and drier, can extend over millennial time scales
hunting), or mass mortality events (Wilmers et al., 2003; (Vereshchagin, 1974; Andrews, 1995; Wambuguh, 2008; Miller,
Ameca y Juárez et al., 2012; Fey et al., 2015; Wenger et al., 2011; Michelutti et al., 2013). In marine ecosystems, the limited
2019). The resulting differences in the abundance, location, and role of bacteria and the temporal stability of environmental
timing of carcass deposition, as well as in animal characteristics conditions can result in slow decomposition rates that foster
including body size and stoichiometry, can have pronounced the development of specialist assemblages on carcasses (Baco
effects on decomposition and utilization of carcass components and Smith, 2003; Smith and Baco, 2003; Beasley et al., 2012).
(Tomberlin and Adler, 1998; Beasley et al., 2012; Subalusky Bone decomposition in aquatic ecosystems may be slowed
and Post, 2018). Carcass decomposition is a multi-stage process: by occasional burial in benthic sediments (Johnston et al.,
an early stage of decomposition characterized by high rates of 2004). Studies suggest only 10–15% of fish bone may be lost
elemental leaching, an active stage characterized by microbial due to permanent burial (Vallentyne, 1960; Schenau and De
and insect colonization, and an advanced stage characterized Lange, 2000; Vanni et al., 2013), although this rate likely varies
by physical/mechanical breakdown and chemical dissolution of widely depending on characteristics of the aquatic ecosystem
bones (Parmenter and Lamarra, 1991; Keenan et al., 2018). The and has not been well-studied. Despite these burial rates,
earlier stages can be relatively rapid, occurring over days to accumulation of detritus from fish bones in marine benthic
months, as compared to the latter stage, as bones can persist for sediments can comprise a significant portion of sediment P and
decades to millennia (Vereshchagin, 1974; Smith and Baco, 2003; lead to high rates of phosphate fluxes under certain conditions
Miller, 2011; Miller et al., 2013). (Schenau and De Lange, 2001).
Much research has focused on the influence of soft tissues, Aquatic ecosystems likely have higher densities of bones than
which provide the majority of carcass resources for invertebrates terrestrial ecosystems because, in addition to mortality of aquatic
and small-bodied vertebrates. Soft tissues are high in nitrogen vertebrates, they may also be a source of mortality for terrestrial
(N) and phosphorus (P), which are often limiting nutrients in animals as well as aggregate slowly-decomposing bones from the
ecosystems. The stoichiometric ratio of N to P in soft tissues terrestrial landscape (Behrensmeyer, 1982; Wenger et al., 2019).
ranges from 10 to 100:1 (Elser et al., 1996). Soft tissues decompose There is a long history of studying the origin and persistence
over days to weeks, but they can have pronounced and rapid of bonebeds and fluvial transport of bones in paleoecology
ecological effects that can persist for long periods of time (Behrensmeyer, 1988, 2007). However, little work has focused on
(Parmenter and Lamarra, 1991; Chaloner et al., 2002; Regester the potential ecological effects of bones in aquatic ecosystems.
and Whiles, 2006; Bump et al., 2009; Parmenter and MacMahon, The disparity in the amount of ecological research on carcasses
2009; Pray et al., 2009). versus that on bones is illustrated in a Web of Science search
Much less research has focused on the decomposition and conducted on 14 March 2019 for literature on the topic. Studies
utilization of bones from animal carcasses (Wambuguh, 2008; on carcass decomposition in aquatic ecosystems [(carcass OR
Wenger et al., 2019). Bone is a composite material consisting body) AND (decomposition OR decay) AND aquatic] since 1990
of a mineral phase, an organic phase, and water (Currey, 2002). yielded 218 studies, as compared to a search for studies on bone
The mineral phase is comprised of calcium phosphate primarily decomposition [(bone OR skeleton) AND (decomposition OR
in the form of hydroxylapatite. The organic phase is comprised decay) AND aquatic] that yielded only 25 studies (Figure 1).
Frontiers in Ecology and Evolution | www.frontiersin.org 2 February 2020 | Volume 8 | Article 31
Subalusky et al. Bone Decomposition Influences Aquatic Ecosystems
FIGURE 1 | Web of Science search conducted on 14 March 2019 for literature since 1990 on decomposition in aquatic ecosystems of carcasses (in gray) versus
bones (in brown).
The Serengeti wildebeest (Connochaetes taurinus) migration population of > 4,000 hippopotamus (Hippopotamus amphibius)
provides an opportunity to examine the influence of large inputs and the annual migration of 1.3 million wildebeest, which both
of bones from large mammals on river ecosystem function, provide important resource subsidies to the river (Subalusky
and raises interesting questions about the ontogeny and effects et al., 2015, 2017). The Serengeti wildebeest migration is in
of animal bones in aquatic ecosystems (Figure 2). Annual the Kenyan portion of the Mara River basin from July to
mass drownings in the Mara River result in the input of an November, and the animals cross the Mara River multiple
average of 6,250 carcasses into the river every year (Subalusky times during this period as they move between dry season
et al., 2017). Approximately half of a wildebeest carcass is feeding grounds. We have documented nearly annual mass
soft tissue, which decomposes over weeks to months, but the drownings of wildebeest during river crossings upstream of the
other half is bone, which comprises 95% of the phosphorus New Mara Bridge near the border between Kenya and Tanzania
(P) in a carcass and decomposes over years (Subalusky et al., (Subalusky et al., 2017).
2017). The pulsed input of these carcasses influences nutrient From 2011 to 2015, we estimated a mean of 6,250 wildebeest
cycling in the river on annual time scales (Subalusky et al., drowned in the river each year, which contributed approximately
2017, 2018), but there may also be long-term effects on 219,200 kg of bones (wet weight) per year to the river (Subalusky
nutrient cycling and river food web dynamics through the et al., 2017). All but one of these drownings occurred within
persistence of bones. a 5 km reach of river, and carcasses tend to accumulate
Here we synthesize several datasets from our research in the on river bends and rock outcroppings within a 5 km reach
Mara River to examine the ecological effects that bone could downstream of the drowning location. Thus, if we assume
have on nutrient cycling, ecosystem function, and food web these bones are distributed along a 10 km reach, and the
structure in the river. We use these data and our preliminary average river width is 40 m, these annual inputs would yield
understanding of the role of bones in this ecosystem to propose an areal density of 0.55 kg bone/m2 . This estimate does
several research questions to improve our broader understanding not account for the continual accrual of bones that occurs
of the role of mammal bones in aquatic ecosystems. We also due to their slow decomposition, and it does not account
suggest this may be an overlooked phenomenon in other rivers for the transport of bones farther downstream that likely
and may have been particularly important in the past when robust occurs over time.
populations of large mammals were more common. All data presented in this paper were from samples collected
just upstream of the New Mara Bridge, which is ∼200 m upstream
of the Tanzanian border, or from an artificial stream experiment
MATERIALS AND METHODS that was conducted inside the Maasai Mara National Reserve. All
wildebeest bones were collected from the carcasses of animals
Study Site that drowned naturally in the river. Fishes were sampled using
This research took place in the Mara River, which runs standard field methods. This study was carried out in accordance
through the Maasai Mara National Reserve in Kenya and with the Yale University Institutional Animal Care and Use
the Serengeti National Park in Tanzania. The river hosts a Committee Animal Use Protocol #2012-10734.
Frontiers in Ecology and Evolution | www.frontiersin.org 3 February 2020 | Volume 8 | Article 31
Subalusky et al. Bone Decomposition Influences Aquatic Ecosystems
FIGURE 2 | The ontogeny of wildebeest bones in the Mara River, Kenya. (A) Annual mass drownings result in the input of a mean of 6,250 carcasses per year.
(B) Carcass soft tissue decomposes over weeks to months, but bone persists in the river for decades. (C) Bones can continue to leach out phosphorus even after a
prolonged period in the river. (D) Biofilms that grow on bones are higher in chl a and organic matter (OM) than biofilms on rocks, and they provide an important food
source for macroinvertebrates and fishes.
Bone Decomposition in days. In Eq. 1, k is the constant decomposition rate of
We measured bone decomposition using three different the material. In Eq. 2, ∝ is the proportion of labile material,
approaches: (1) measuring in situ mass loss of bones in litterbags k1 is the decomposition rate of labile material, and k2 is the
in the river, (2) measuring changes in the elemental composition decomposition rate of recalcitrant material. All models were run
of bones after an extended time in the river, and (3) measuring for 500 iterations. We used the resulting parameter values for the
nutrient leaching rates from bones in microcosms. selected model to estimate time to 95% mass loss as ln(0.05)/k for
First, we placed samples of four different types of fresh bone the labile and recalcitrant components.
(triplicate samples of leg, rib, scapula, vertebrae; n = 12) inside We also used these parameter values to calculate the steady-
fine mesh (
Subalusky et al. Bone Decomposition Influences Aquatic Ecosystems
United States). The percent organic matter (OM) was measured which is comparable to our estimates of areal density of bones
by weighing samples before and after combustion. We used t-tests in the Mara River.
to compare the % OM, % C, % N, and % P in fresh bones versus We used a Manta2 sonde (Eureka Environmental, Austin, TX,
aged bones of different types (R Core Team, 2018). We compared United States) containing a Cyclops-7F submersible fluorometer
fresh rib, vertebrae, and joint bones to aged rib, vertebrae, and leg (Turner Designs, San Jose, CA, United States) to measure water
bones, respectively. column chlorophyll a (chl a) weekly. We collected water samples
Third, we measured initial leaching rates of bone by placing weekly to analyze inorganic nutrients, as described above. We
sub-samples of fresh bone (66–98 g, mean = 80 g) in chambers also destructively sampled one ceramic tile each week to measure
(n = 3) that were filled with 4 L of unfiltered river water and OM of the biofilm as ash free dry mass (AFDM). We filtered a
open to the environment. We collected 50 mL water samples known volume of sample through a pre-weighed, pre-combusted
for analysis of inorganic nutrients every ∼6 days for 31 days. Whatman GF/F filter (GE Healthcare Bio-Sciences, Pittsburgh,
Water samples were collected using a syringe and filtered PA, United States), and measured AFDM by drying the filter at
through a 0.2 µm Supor polysulfone filter (Pall Corporation, Port 60◦ C, re-weighing it, combusting it for 4 h at 450◦ C and then
Washington, NY, United States) into a sample bottle and frozen re-weighing it to determine mass loss upon combustion. In the
until analysis. Samples were analyzed on a portable flow injection final week of the experiment, we scrubbed the biofilm off one
analyzer in the field. Ammonium was analyzed using the gas ceramic tile from each stream and measured the concentration
exchange method (APHA, 2006). Nitrate was analyzed using zinc of chl a in a known volume of water using the Manta2 sonde,
reduction (Ellis et al., 2011). Soluble reactive phosphate (SRP) which we then converted to chl a per unit area of tile. In situ
was analyzed using the molybdate blue method (APHA, 2006). chl a fluorescence can be used as a proxy for chl a concentration,
Nutrient concentrations were multiplied by the volume of water although it may provide an overestimate, and direct comparison
in the chamber at each sampling time point to obtain total mass with chl a concentrations requires calibration (Holm-Hansen
of nutrients leached. We did not correct for background nutrient et al., 2000; Roesler et al., 2017). However, in this analysis, we only
concentrations in the water we used to fill the chambers, as we compared in situ fluorescence values across treatments.
did not maintain control chambers over time, but concentrations Data were analyzed for normality using a Shapiro–Wilk’s
were very low compared to leaching rates of ammonium and SRP. normality test in R Core Team (2018), and water column
The mass of the bone sample was multiplied by the % N and % P nutrients, water column chl a, and benthic AFDM were log-
measured for fresh joint bones, and the ammonium and SRP that transformed to meet statistical assumptions. We examined
leached out over 31 days was measured as a proportion of the differences in water column nutrients, water column chl a, and
total N and P in the bone sample. benthic AFDM throughout the duration of the experiment using
a linear mixed-effect model run with the lme function in the
nlme package in Pinheiro et al. (2016) and R Core Team (2018).
Effects of Bones on Ecosystem Function We fitted lme models with the restricted maximum likelihood
in Experimental Streams method and a continuous autoregressive temporal correlation
As part of a larger mesocosm experiment examining the influence structure with week as the repeated factor. Treatment (control,
of wildlife subsidies on ecosystem function, we used recirculating bone) and time (each of 3 weeks of measurement) were treated
experimental streams to compare the influence of bone versus as fixed effects, and individual streams were treated as random
rock substrates on water column nutrient concentrations, effects. We then used the lsmeans package to perform a Tukey
and water column and benthic production. Details of the pairwise comparison test between treatments for parameters over
experimental stream array are in Subalusky et al. (2018). In the duration of the experiment (Lenth, 2016; R Core Team, 2018).
this experiment, we had 18 individual streams (three blocks We also analyzed the effect of treatment on biofilm chl a at the
of six streams each), and treatments were randomly assigned end of the experiment with a one-way ANOVA using the aov
among each block. The full experiment included controls (n = 4), function in R Core Team (2018).
and four different treatments (n = 2–4). We only present here
data from the control streams (n = 4) and the bone treatment Bone Biofilm
streams (n = 2). We analyzed chl a and OM (measured as AFDM) of biofilms
One 5-L bucket of washed gravel was placed along the bottom collected from both wildebeest bones and rocks in the Mara River
of each stream channel as substrate, and five ceramic tiles in November 2013 (during the wet season) and February 2014
were placed in the channel bed for sampling. Streams were (during the dry season). At both sampling times, we selected
filled with 60 L water from Emarti Bridge, which is on the three rocks and three bones from the same reach of river,
Mara River upstream of the influence of large wildlife, and scrubbed the entire upper surface of the substrate clean using a
inoculated with periphyton scrubbed from rocks from New Mara toothbrush, and analyzed photos of the substrates using ImageJ
Bridge, within the range of wildlife. Streams were allowed to software to measure the surface area (Schneider et al., 2012).
equilibrate for 1 week, after which treatments were applied, and We filtered a known volume of sample through a pre-weighed,
the experiment was run for two additional weeks. In the bone pre-combusted Whatman GF/F filter, and measured AFDM as
treatment streams, half of the volume of gravel was replaced with described above. We filtered a known volume of sample through
wildebeest bones of unknown age that had been removed from a second Whatman GF/F filter for analysis of chl a. We froze the
the river. This treatment had approximately 0.7 kg bones m−2 , filter paper for > 24 h, extracted the chl a using methanol with
Frontiers in Ecology and Evolution | www.frontiersin.org 5 February 2020 | Volume 8 | Article 31
Subalusky et al. Bone Decomposition Influences Aquatic Ecosystems
a basic pH (Holm-Hansen, 1978), and analyzed the samples on We collected 16–30 individuals from each of four families of
a Turner Aquafluor handheld fluorometer (Turner Designs, San aquatic macroinvertebrates, including Baetidae, Hydropsychidae,
Jose, CA, United States). We calculated both chl a and AFDM per Caenidae, and Simulidae, and we combined individuals into a
unit surface area of the substrate. Data were tested for normality single bulk sample per family. For Baetidae and Simulidae, we
using a Shapiro–Wilk’s normality test in R, and we analyzed the had sufficient individuals to run two replicate samples of 30
effect of substrate and season on both parameters using a two- individuals each, and we used the mean of those replicates for
way ANOVA with the aov function followed by a Tukey HSD test the stable isotope signatures of those taxa. We also collected
in R Core Team (2018). tissue samples from the lateral muscle of three species of fishes,
We analyzed community composition of biofilms from both including Labeo victorianus (n = 8), Labeobarbus altianalis
wildebeest bones and rocks collected from the Mara River (n = 5), and Bagrus docmac (n = 3). All samples were collected
in October 2017 and November 2018. In 2017, we scrubbed from near the New Mara Bridge. All samples were dried,
biofilms from the surfaces of three bones and three rocks ground into a fine powder, and analyzed for δ13 C and δ15 N
randomly selected from the same reach of river, although on a ThermoFinnigan Delta Plus Advantage stable-isotope mass
sampling was not done quantitatively. Samples were preserved spectrometer (Thermo Scientific, Boca Raton, FL, United States)
with Lugol’s solution and counted in the lab at 400x on a coupled to a Costech ECS 4010 Elemental Analyzer (Costech
Leica DM LS2 compound microscope until 100 algal cells Analytical Technologies, Inc., Valencia, CA, United States).
had been reached, and abundance of each taxa was given as We used Bayesian mixing models in MixSIAR to
a proportion of the total. In 2018, we collected three bones estimate the proportion of each basal resource assimilated
and three rocks from the same reach of river, making sure in macroinvertebrate and fish tissue (Moore and Semmens, 2008;
to collect paired samples from similar depths, and scrubbed Stock and Semmens, 2013). The results of fish assimilation were
16 cm2 of surface area. Samples were again preserved with Lugol’s analyzed and presented by species in Subalusky et al. (2017);
solution and counted in the lab. We counted 10 microscope here, we analyzed assimilation across aquatic macroinvertebrates
fields for each sample, and we calculated the total abundance and fishes as composite consumer groups. All fish species were
of each taxa. We identified both bone and rock periphyton to omnivorous, so we used 0.4 ± 1.3 for δ13 C (Post, 2002) and
phylum (Chlorophyta, Chrysophyta, Cyanobacteria, Euglenozoa) 4.3 ± 1.5 for δ15 N (Bunn et al., 2013) for fish trophic enrichment
(Prescott, 1978), and we parsed Chlorophyta into three functional factors, which incorporates variability in trophic structure. We
groups based on growth form (unicellular, colonial, and used 0.4 ± 1.3 for δ13 C (Post, 2002) and 1.4 ± 1.4 for δ15 N (Bunn
filamentous). We conducted an analysis of similarity on the et al., 2013) for macroinvertebrate trophic enrichment factors.
community data separately for each year using the anosim We ran models with the normal MCMC parameters (100,000
function in the vegan package in R Core Team (2018) and chain length, 50,000 burn-in). Visual analysis of isospace plots
Oksanen et al. (2019). The function vegdist is used to create confirmed that consumer data were within the minimum convex
a Bray dissimilarity matrix, and anosim uses the rank order of polygon of source data, suggesting we were not missing any
dissimilarity values to test for statistically significant differences major diet sources (Phillips et al., 2014).
between communities.
Stable Isotopes RESULTS
We used C and N stable isotopes to examine isotopic differences
between biofilms on rocks and biofilms on bones over three Bone Decomposition
different seasons. We collected biofilms from rocks and bones The in situ decomposition of bone in the Mara River was much
in November 2013 (wet season), February 2014 (dry season), better described by the two-pool model of decomposition than
and July 2016 (wet season) (n = 3 of each type in each by the single-pool model for all four bone types (Table 1).
season). We also analyzed the stable isotope signature of fresh Results from this model suggest different bone types vary
wildebeest bones (n = 8) collected in 2012–2013 to help interpret in their decomposition rate. Bones are comprised of 7–27%
differences in biofilm signature between bones and rocks. We labile material that decomposes over 78–119 days, and 73–93%
then used C and N stable isotopes to partition the contribution refractory material that decomposes over > 80 years (Table 1).
of various basal food web resources to the tissue assimilation of The k value for the refractory material in all bone types
aquatic macroinvertebrates and fishes. We used sample data from reached the minimum bounds in this analysis package (0.0001),
February 2014, as this was 4 months after any fresh wildebeest providing a minimum estimate for the time to 95% loss; however,
carcasses were in the river. This time period should exceed extrapolation beyond this time point is well outside the bounds of
the typical isotope turnover rate for consumer muscle tissue what we can infer with the relatively limited duration of our field
(Vander Zanden et al., 2015) and thus minimize the signal of data (216 days). Scapula and leg bones had the lowest proportion
wildebeest carcass soft tissue in the consumers. We used biofilms of labile material and as a result decomposed the most slowly
growing on rocks and on bones to characterize autochthonous (Figure 3). Vertebrae bones had the highest proportion of labile
basal food web resources, and we collected samples of hippo material and decomposed more quickly than the other bone
feces (n = 9), which is the primary source of allochthonous food types. It is unlikely that bone mass loss during this experiment
web resources in this region of the river (Masese et al., 2015; was due to downstream transport of particulate material, due
Subalusky et al., 2015, 2018). to the fine mesh size of the litterbags used. Based on an annual
Frontiers in Ecology and Evolution | www.frontiersin.org 6 February 2020 | Volume 8 | Article 31
Subalusky et al. Bone Decomposition Influences Aquatic Ecosystems
input of 219,200 kg (wet mass), we estimate the steady-state
TABLE 1 | Comparison of single- and dual-pool negative exponential models of decomposition for four different wildebeest bone types, parameter values for the best fit model (AIC values in bold; the dual-pool model
was the best fit for all bone types), estimates of time to 95% loss for labile and recalcitrant material, and overall% organic matter (OM) and C:N:P stoichiometry by percent mass for wildebeest bones after 216 days in
14:4:11
16:3:12
15:4:11
23:4:10
14:4:11
(13,1,2)
38:3:8
C:N:P
Aged
standing stock of bones in the Mara River is 4.4 × 106 to
5.6 × 106 kg. This estimate is likely high, as it assumes the system
is in equilibrium, and it is based on a conservative decay rate due
25:4:10
18:5:11
22:4:10
23:5:9
to microbial decomposition that does not account for loss from
C:N:P
(4,1,1)
Fresh
animal consumption or mechanical breakdown.
There were relatively small differences in the stoichiometry
of fresh versus aged bones (those that had been in the river for
37 (16)
% OM
Aged
33
28
27
54
37
216 days) (Figure 4 and Table 1). The mean % OM decreased
on average from 42 ± 11% in fresh bones to 37 ± 16% in aged
bones. The average stoichiometry of fresh bones (joint, rib, and
42 (11)
% OM
Fresh
vertebrae) was 22.1 C: 4.5 N: 9.9 P by % mass compared to 23.1
48
37
41
C: 3.5 N: 10.2 P for aged bones. In leg and rib bones, the % C
and % N declined, while the % P increased, likely due to the
Standing stock
relatively higher % C and N of labile material in bone (e.g., lipids)
(×106 kg)
and the higher % P of refractory material (e.g., apatite). However,
5.6
5.5
5.0
4.4
in vertebrae, % C increased as % N and % P decreased over time,
which may reflect a greater degree of vascularization and greater
initial proportion of labile N and P in this bone type. The only
significant changes were the decrease in % N in rib bones (t-
Time to 95%
loss (years)
test, t = −6.53 .2 , p = 0.006) and the increase in % P in leg bones
*Fresh joint bone was compared to aged leg bone.**Data from a 3360 year BP (14 C age) bison skull found in Clear Lake, IA, United States.
>82
>82
>82
>82
(t = −4.253 .9 , p = 0.014) (Figure 4).
In the chamber experiment, approximately 50% of the mass
Recalcitrant
of SRP and ammonium that leached out of the bones over a
Proportion k2 (day−1 )
month was available after only 3 days (Figure 5). The mass of
0.0001
0.0001
0.0001
0.0001
ammonium that leached out after 1 month (96.8 ± 35.7 mg,
mean ± SD) was almost an order of magnitude larger than that of
SRP (12.7 ± 2.9 mg). Background values from the water used in
0.9253
0.8315
0.7328
the chambers (SRP = 0.16 mg, NH4 = 0.50 mg) were ∼1% of the
0.911
final values. Ammonium appeared to stabilize during the latter
half of the month, which may have been due to equilibration
with the atmosphere, while SRP continued to increase. A large
Time to 95%
loss (days)
amount of nitrate available on day 1 (5.0 ± 0.3 mg) was due to
111
109
119
78
the water used in the chamber, which had a background nitrate
value of 4.8 mg, but nitrate levels fell to nearly zero by day 3
and stayed there for the duration of the study. This decline was
Labile
Dual-pool Proportion k1 (day−1 )
likely due to loss through denitrification due to anoxic conditions
0.0271
0.0382
0.0275
0.0251
in the mesocosms, which we did not measure. We also did not
measure other forms of nutrient uptake that may have occurred
in these chambers; thus, our estimates of leaching rates are likely
0.0747
0.0890
0.1685
0.2672
conservative. After 31 days, we estimate the bone samples leached
out 3.2 ± 0.7% (mean ± SD) of the initial N as ammonium and
the river and a bison skull after 1000s of years underwater.
0.2 ± 0.0% of the initial P as SRP.
−38.13
−36.14
−34.04
−39.75
model
Effects of Bones on Ecosystem Function
in Experimental Streams
AIC
Single-pool
There was a significant effect of both treatment (LME ANOVA:
−14.26
−21.83
−17.99
−14.68
model
F 5 ,1 = 213.621, p < 0.001) and time (F 5 ,2 = 22.547, p < 0.001),
and a significant interaction between them (F 5 ,2 = 81.530,
p < 0.001), on water column SRP in the experimental streams
(Figure 6A). There was no difference between the bone treatment
and the control in week 1, before treatments were applied,
indicating similar background conditions. After the treatments
Bone type
were applied, the bone treatment had > 300 times higher SRP
Vertebrae
Scapula
Bison**
than the control treatment in week 2 (p < 0.001) and 150 times
Mean
Leg*
(SD)
Rib
higher in week 3 (p = 0.001) of the experiment.
Frontiers in Ecology and Evolution | www.frontiersin.org 7 February 2020 | Volume 8 | Article 31
Subalusky et al. Bone Decomposition Influences Aquatic Ecosystems
FIGURE 3 | Mean (± SE) proportion of biomass remaining for scapula, leg, rib, and vertebrae bones (n = 3 per bone type) from a wildebeest carcass during litterbag
deployment in the Mara River with best fit models following a parallel discrete model of decomposition.
FIGURE 4 | Percent (A) organic matter, (B) carbon, (C) nitrogen, and (D) phosphorus in leg, rib, and vertebrae bones (n = 3 per bone type) from wildebeest when
fresh and after 216 days in the Mara River. Fresh joint bone was compared to aged leg bone. Asterisks indicate significant difference, where *p < 0.05 and
**p < 0.01.
There was no significant effect of treatment or time, statistically significant (p = 0.0737), likely due to low replication
or significant interactions between them, for ammonium (Figure 7A). There was no difference between the bone and
(Figure 6B). There was a significant effect of time on NO3 control treatment in week 1 or week 3.
(F 5 ,2 = 204.075, p < 0.001) although no treatment effect, as both There was no significant effect of treatment or time, or a
the bone and control treatment declined from ∼600 µg L−1 NO3 significant interaction between them, on tile biofilm AFDM
to nearly zero between weeks 1 and 2 (Figure 6C). (Figure 7B). There also was no significant effect of treatment on
There was no significant effect of treatment on water column tile biofilm chl a at the end of the artificial stream experiment
chl a in the experimental streams, but there was a significant effect (ANOVA: F 5 ,1 = 0.171, p = 0.7).
of time (LME ANOVA: F 5 ,2 = 9.972, p = 0.007) and a significant
interaction between treatment and time (F 5 ,2 = 17.369, p = 0.001). Bone Biofilm
Chl a was 5 times higher in the bone treatment (364 ± 32) than Biofilm on bones had significantly higher chl a (two-way
the control treatment 70 ± 17) in week 2, although this was not ANOVA; F 1 ,11 = 14.64, p = 0.005) and OM (two-way ANOVA;
Frontiers in Ecology and Evolution | www.frontiersin.org 8 February 2020 | Volume 8 | Article 31
Subalusky et al. Bone Decomposition Influences Aquatic Ecosystems FIGURE 5 | Mean (± SE) total mass in 5-L microcosms (n = 3) filled with river water of (A) soluble reactive phosphorus, (B) ammonium, and (C) nitrate leached out of wildebeest leg bone. FIGURE 6 | Water column (A) soluble reactive phosphorus, (B) ammonium, and (C) nitrate in experimental streams with all gravel benthos (control treatment; n = 4) or half gravel-half bone benthos (bone treatment; n = 2) in Week 1, after equilibration and just before treatments were applied, and in Weeks 2 and 3 of the experiment. Asterisks indicate significant difference, where **p < 0.01 and ***p < 0.001. F 1 ,11 = 9.13, p = 0.017) than biofilms on rocks (Figure 8). Chl a There was no significant difference between the algal was 4.6 times higher and AFDM was 2.0 higher on bone biofilm communities in biofilms growing on bones and those growing than on rock biofilm. There was no significant effect of season or on rocks in either year (2017: ANOSIM R = −0.1481, p = 0.9; interaction between season and substrate. 2018: ANOSIM R = −0.1852, p = 0.8). The negative R values Frontiers in Ecology and Evolution | www.frontiersin.org 9 February 2020 | Volume 8 | Article 31
Subalusky et al. Bone Decomposition Influences Aquatic Ecosystems
FIGURE 7 | (A) Water column chl a and (B) biofilm ash free dry mass (AFDM) on ceramic tiles in experimental streams with all gravel benthos (control treatment;
n = 4) or half gravel-half bone benthos (bone treatment; n = 2) in Week 1, after equilibration and just before treatments were applied, and in Weeks 2 and 3 of the
experiment, after treatments were applied.
FIGURE 8 | Mean (± SE) values of (A) chlorophyll a and (B) organic matter as ash free dry mass (AFDM) on bone (n = 3) and rock (n = 3) substrates in the Mara
River. Asterisks indicate significant difference, where *p < 0.05 and **p < 0.01.
indicate greater dissimilarities within groups than between (4–10h) (Figure 10). The δ13 C of biofilm on bones was much
groups (Figure 9). In 2017, unicellular algae were the most closer to the δ13 C of wildebeest bones themselves (1.6–3.7h
common, followed by colonial algae and diatoms. In 2018, difference). These data suggest biofilms on both rocks and bones
diatoms were the most common, followed by filamentous may be obtaining N from the water column, but biofilms on
and colonial algae. bones may be obtaining some C from the bones themselves.
Sufficient differences between the three dominant basal
Stable Isotopes resources at NMB in February 2014 (bone biofilm, rock biofilm,
Biofilm on bones had a δ15 N similar to that of biofilm on rocks and CPOM) enabled us to parse their influence on aquatic
(0.4–1.7h difference) but a δ13 C that was much more enriched consumers (Table 2). Mixing model results showed that bone
Frontiers in Ecology and Evolution | www.frontiersin.org 10 February 2020 | Volume 8 | Article 31Subalusky et al. Bone Decomposition Influences Aquatic Ecosystems FIGURE 9 | Algal community composition on bone (n = 3) and rock (n = 3) substrates in the Mara River, where individuals are presented as (A) a proportion of 100 individuals counted in 2017, or (B) total number within 10 microscope fields in 2018. Green algae are parsed by growth form in the inset figures. FIGURE 10 | C and N stable isotope signatures of wildebeest bone (n = 8) and biofilm on rock and on bone on three sampling dates (n = 3 of each at each time point) in the Mara River. Frontiers in Ecology and Evolution | www.frontiersin.org 11 February 2020 | Volume 8 | Article 31
Subalusky et al. Bone Decomposition Influences Aquatic Ecosystems
TABLE 2 | Stable isotope signatures of basal food web resources, aquatic greater quantity and/or quality of biofilms. Increased biofilm
macroinvertebrates, and fishes collected in February 2014, at the New Mara
growth on bones could be due to leaching of nutrients from the
Bridge in the Mara River.
bones and/or greater surface roughness of bones as compared
Category n δ13 C (mean ± SD) δ15 N (mean ± SD) to rocks, which could provide increased surface area for growth
(Bergey and Cooper, 2015). There did not appear to be a
Hippo feces 9 −14.7 ± 1.0 3.8 ± 1.0
difference in algal composition on bones versus rocks (Figure 9).
Biofilms on rocks 3 −19.2 ± 0.6 13.0 ± 1.2
The isotopic signature varied between biofilms on bones and
Biofilms on bones 3 −15.2 ± 2.3 11.2 ± 1.8
those on rocks, with the δ13 C of bone biofilm being closer
Aquatic macroinvertebrates 4* −15.8 ± 1.8 7.8 ± 1.7
to that of wildebeest bones themselves (Figure 10). These
Fishes 16 −13.8 ± 1.1 10.2 ± 0.8
data suggest some biofilm constituents may be able to utilize
*Each macroinvertebrate sample represents one family comprised of a composite elements leaching from the bones, particularly C, which could
sample of ≤ 30 individuals.
have implications for bone biofilm quantity and quality. Mixing
model analysis suggests aquatic macroinvertebrates and fishes
biofilm accounts for 19 ± 16% (mean ± SD) of assimilated assimilate a similar proportion of bone biofilm, which is equal
tissue in macroinvertebrates and 24 ± 13% in fishes in the to or greater than the proportion of rock biofilm assimilated
Mara River (Figure 11). These proportions are similar to that (Figure 11). Given the likely much lower abundance of bones
of rock biofilm for macroinvertebrates (21 ± 13%), but greater on the river bottom as compared to rocks, these data suggest
than rock biofilm for fishes (7 ± 6%). The remaining portion of consumers may be preferentially feeding on biofilms growing on
both macroinvertebrate and fish diet is comprised of hippo feces, bones. If so, that preference could be driven by higher density or
which accounts for nearly all of the particulate OM at this site. quality of the biofilms growing on bones as compared to those
The results support an important contribution of bone; however, growing on rocks.
we note that the 95% Bayesian credible intervals are quite wide Many of these analyses are based on small sample sizes, and
and overlap for most of the resources, indicating a reasonable we synthesize them here to stimulate areas for future research.
amount of uncertainty in the contribution. Altogether, these data suggest bones may play an important
and persistent ecological role in the Mara River, and they raise
several over-arching questions about the potential ecological role
DISCUSSION of bones in aquatic ecosystems in general.
Wildebeest bones provide a distinct and complex resource in
the Mara River, and given their abundance in this system, they
What Is the Magnitude and Frequency of
may influence nutrient cycling, ecosystem production, and food Animal Bone Inputs to Aquatic
web dynamics at the river ecosystem scale. Wildebeest bone is Ecosystems?
comprised of 22% C, 4% N, and 10% P, although stoichiometry The Mara River receives annual inputs of large mammal bones,
and decomposition rates vary by bone type (Table 1). A mean and bones persist in this system over long timescales. The large
of 15% of the initial mass of bones is relatively labile and input rate and slow decomposition rate yield a steady-state
decomposes over 80–120 days (Figure 3). Initial leaching releases standing stock estimate of 4.4 × 106 to 5.6 × 106 kg of bones
a large amount of inorganic N relative to P (Figure 5). After this in the river, which is equivalent to the biomass of 49 blue whales.
labile portion is gone, the more refractory material that is rich in This estimate is likely high, as it assumes that the system is in
P can persist in the system for decades, although mineralization equilibrium and input rates have been constant over time, and
and leaching of P continue to occur. For example, bones that it is based on a conservative decomposition rate that does not
had been submersed in the river for an unknown period of account for animal consumption or mechanical breakdown. This
time still released SRP when placed in the experimental stream estimate also does not reflect the potential distribution of bones
channels (Figure 6). The increase in SRP during week 2 of the downstream through the river system and into the Mara Wetland
experiment may have reflected either a pulse of SRP release and Lake Victoria. Nevertheless, it reflects the magnitude of bone
in response to changing environmental conditions between the that can accrue in a system that receives large inputs of carcasses,
river and the experimental streams, or a steady release of SRP particularly of large mammals. How typical is this of other aquatic
that declined in week 3 due to uptake. These data suggest that ecosystems?
large mammal bones play a unique role as slow-release nutrient Animal bones can enter aquatic ecosystems through various
subsidies in this system. pathways. Wenger et al. (2019) proposed a framework for
Nutrient leaching from bones may stimulate increased classifying animal carcass inputs to aquatic ecosystems, in which
production. In the experimental stream channels, the increase in inputs may be either autochthonous and continuous (e.g., annual
SRP in the bone treatment in week 2 was coincident with five mortality from aquatic animals), autochthonous and pulsed
times more water column chl a but no change in tile AFDM (e.g., mass mortality of aquatic animals), allochthonous and
(Figure 7), suggesting P from bones stimulated water column continuous (e.g., annual mortality from terrestrial animals that
primary production. In the river, biofilms growing on bones had are transported in from the watershed), or allochthonous and
five times more chl a and two times more OM than biofilms pulsed (e.g., seasonal mortality from terrestrial animals that
growing on rocks (Figure 8), suggesting bones may support perish in situ). The wildebeest inputs to the Mara River represent
Frontiers in Ecology and Evolution | www.frontiersin.org 12 February 2020 | Volume 8 | Article 31Subalusky et al. Bone Decomposition Influences Aquatic Ecosystems
FIGURE 11 | Proportion of basal resources in assimilated tissue (mean and 95% Bayesian credible intervals) of (A) aquatic macroinvertebrates (n = 4 taxa,
composite samples of 16–30 individuals/taxa) and (B) fishes (n = 3 taxa, 3–8 individuals/taxa) in February 2014 in the Mara River from hippo feces (n = 9), bone
biofilm (n = 3), and rock biofilm (n = 3).
an example of allochthonous and pulsed, which has the potential that the recalcitrant nutrients can continue to leach out of bone
to be one of the largest sources of inputs in certain systems. at longer time scales. Thus, bones have the ability to provide
For example, research building on historical, anecdotal accounts a slow-release nutrient subsidy to aquatic ecosystems, which
suggests that large-scale inputs of bison bones may have been lengthens the temporal scale at which we normally consider
commonplace in the rivers of the American Great Plains as animal influences on nutrient cycling. What are the rates of
recently as the late 1700s. A mass drowning of bison in the P mineralization from the recalcitrant portion of bones, and
Assiniboine River could have comprised ∼50% of the annual P what other elements may be leaching from the bones, such
load for that river (Saindon, 2003; Wenger et al., 2019). It is as calcium? How do these mineralization rates change across
unknown how the magnitude of these inputs would compare environmental conditions and over time? And how may the
to those resulting from in situ mortality of aquatic vertebrates, attachment of algae and microbes facilitate the erosion of bones
such as fishes. However, allochthonous inputs from terrestrial through alteration of the boundary layer pH or scavenging
animals that transport additional resources into the system are of minerals? Much of the forensic and archeological study of
likely to have different and perhaps more pronounced ecosystem bone decomposition has focused on bones buried in soil, and
effects as compared to autochthonous ones (Subalusky and Post, research suggests increasing soil temperature, moisture content
2018). Furthermore, the size of bones likely influences their and geochemistry are all important variables in driving microbial
stoichiometry and decomposition rate, such that larger bones decomposition, although a great deal of variability occurs in
may be expected to provide the slow release of nutrients we bones even within the same site (Hedges, 2002; Christensen
observed with wildebeest bones, whereas smaller bones may and Myers, 2011; High et al., 2015). In aquatic ecosystems,
decompose more quickly (Nobre et al., 2019). Much work where less research has been done on bone decomposition, the
remains to be done to quantify the magnitude of animal carcass decomposition process can be even more variable due to the large
inputs from these four categories across ecosystems and over time number of variables that can influence the process, including
and space. We expect that rates of allochthonous carcass inputs to temperature, water depth, currents, dissolved oxygen, dissolved
aquatic ecosystems would be highest in higher order rivers, which OM, substrate type, salinity, acidity, and insect and scavenging
both aggregate more from the watershed and may provide a cause activity (Simon et al., 1994; Heaton et al., 2010). In lake and
of direct mortality for animals, and in landscapes with abundant river ecosystems, bone burial in benthic sediments is likely to
populations of large mammals and particularly migratory herds. slow or stop decomposition. Studies of fish bones suggest 10–
15% of bones may be permanently buried in lakebed sediments
(Vallentyne, 1960; Schenau and De Lange, 2000; Vanni et al.,
How Bioavailable Are the Elements in 2013). How do decomposition processes vary across species and
Bones, and Over What Time Scales Do ecosystem types, and how bioavailable are elements throughout
They Become Available? these processes?
Our data suggest most of the labile nutrients leach out of bone As part of this study, we analyzed the C, N, and P composition
within a few months of deposition; however, they also indicate of a bison skull that was recently recovered from Clear Lake, IA,
Frontiers in Ecology and Evolution | www.frontiersin.org 13 February 2020 | Volume 8 | Article 31Subalusky et al. Bone Decomposition Influences Aquatic Ecosystems
United States, in order to understand how extended submersion What Comprises the “Osteophyton”
in an aquatic ecosystem (e.g., over centuries to millennia) Community, or the Biofilm Community
could influence bone composition in another species. Modern
American bison (Bison bison) have been extirpated from that That Grows on Bones?
region for several 100 years; however, preliminary analysis based In this study we analyzed the algal community at a coarse
on the skull’s shape and size suggests this may be a prehistoric taxonomic resolution and found no differences between biofilms
specimen (Skinner and Kaisen, 1947). We conducted radiocarbon growing on bones and rocks, but do differences occur at a lower
dating to determine the skull is 3,360 ± 25 years BP (14 C age); algal taxonomic resolution, or in the bacterial community? There
thus, this skull could have been in the water for 1000s of years. are biofilm organisms, referred to as epibionts, that appear to
The bison skull had very similar proportion of OM (37%) and specialize in colonizing surfaces of aquatic flora and fauna. For
C:N:P stoichiometry (13.7 C: 4.3 N: 10.6 P by % mass) to the example, certain taxa of filamentous algae and diatoms have
fresh rib and aged scapula bones from the wildebeest, suggesting been found to specialize on turtle shells, likely due to the ability
that even bones 100s to 1000s of years old may still retain a fairly of those taxa to withstand frequent wetting and drying periods
large proportion of organic material. In the case of the bison skull, (Edgreen et al., 1953; Wu and Bergey, 2017). In some cases,
burial in lake sediments in a cold region likely maintained its the primary production of these epibionts may alter the net
relatively intact condition. metabolism of their host (Lukens et al., 2017). Are there taxa that
specialize on bones, either due to greater surface area roughness
that enables better attachment or to their ability to utilize carbon
Are Animal Bones Capable of Influencing or other elements leaching from the bones? Do these taxa provide
Aquatic Ecosystem Function? a higher quality food resource for grazers? Are there aquatic
Altogether, our data suggest bones may provide important species that directly consume bones, similar to rodents and
nutrient and microhabitat resources at local scales in this system. hyenas in terrestrial ecosystems? Crocodilians can digest bones
However, it is still unclear to what degree these effects scale when consuming entire carcasses (Fisher, 1981). It is unclear
up to influence the whole river ecosystem. To what degree whether bones would be directly consumed as a food resource,
can long-term mineralization of bone inputs support primary, due to their relatively low caloric content, but they may provide
and ultimately secondary, production in aquatic ecosystems? other elements, such as calcium, that are otherwise limiting in
The answer likely depends upon both the magnitude of the the system.
inputs and the environmental context of the aquatic ecosystem
(Subalusky and Post, 2018). In the Mara River, wildebeest bones
contribute a substantial portion of the annual P flux from CONCLUSION
upstream (Subalusky et al., 2017). P flux is significantly higher
at the site where wildebeest drownings occur, but only when Wildebeest bones appear to play an important ecological role
carcasses are present, suggesting long-term leaching of bones in the Mara River system due to the quantity and frequency
does not significantly elevate P flux at the site scale (Subalusky of input and their potential to influence short- and long-term
et al., 2018). However, it may be readily assimilated and stimulate nutrient cycling, production and aquatic food webs. Animal
primary production, as we observed in the experimental streams. bones may play an important role in other aquatic ecosystems.
Indeed, water column chl a is higher at the site where drownings In many ways, the role of bones may be similar to that of coarse
occur, although the degree to which that production is due to woody debris in some systems, as both decompose over long
wildebeest inputs versus other drivers is still unclear (Subalusky time scales, provide food and structural habitat, and are capable
et al., 2018). There also may be more localized hotspots of P of entraining finer particulates (Wohl, 2017). The role of bones
availability by bone piles that could have ecological consequences. may have been even more important before declines in animal
Whole ecosystem effects of wildebeest bones may be muted populations, and declines in large mammals and migratory herds
in the Mara River, where even larger resource subsidies from in particular, reduced the occurrence of large animals on the
hippos have pronounced ecosystem effects (Subalusky et al., landscape. However, domestic animals may be replacing that role
2018). However, in systems with substantial inputs of large in some places. The biomass of domestic livestock and poultry
bones and low background nutrient loading, it is possible that now far exceeds that of wild mammals and birds (Bar-On et al.,
bones could influence ecosystem function over long time scales. 2018). Most livestock carcasses are fully used or disposed of in
For example, in a study of ponds near early Arctic hunting controlled ways, and bones are often used as fertilizer or animal
communities, marine mammal bones from butchered carcasses feed (Jayathilakan et al., 2012). However, catastrophic flooding,
were still evident in those systems 500- > 1500 years later, due to which has become increasingly common in some regions due to
slow decomposition rates, and they still influenced pond water climate change, can lead to mass mortality of domestic livestock
nutrients and production (Michelutti et al., 2013). Bones also and transport of livestock carcasses into aquatic ecosystems.
may play an important role in providing structural habitat in Animal bones from annual mortality may play a relatively small
rivers that otherwise lack it. Although this is not the case in the role in most aquatic ecosystems, but pulsed inputs from mass
Mara River, which has a large degree of rock and boulder cover, mortality events could be a substantial component of nutrient
it may be an important role of bones in other rivers flowing budgets (Wenger et al., 2019), and the bones could persist for
through grasslands. decades or longer in the system. The role of bones in aquatic
Frontiers in Ecology and Evolution | www.frontiersin.org 14 February 2020 | Volume 8 | Article 31Subalusky et al. Bone Decomposition Influences Aquatic Ecosystems
ecosystems is an area that deserves more study given the unique FUNDING
and long-lasting influence bones may have.
Funding was provided by US National Science Foundation
grants to DP (DEB 1354053 and 1753727) and ER (DEB
DATA AVAILABILITY STATEMENT 1354062), a grant from the National Geographic Society to
DP, and a fellowship from the Robert and Patricia Switzer
The datasets analyzed for this study and all code for the statistical Foundation to AS.
analyses and figures are included in the Supplementary Material.
ACKNOWLEDGMENTS
ETHICS STATEMENT
We thank Brian Heath and the Mara Conservancy for support
This study was carried out in accordance with the Yale University with our field research; Paul Geemi, James Landefeld, and Ella
Institutional Animal Care and Use Committee Animal Use Bayer for assistance in the field and with the artificial stream
Protocol #2012-10734. This research was conducted under a experiment; Scott Grummer of the Iowa DNR Clear Lake office
research permit from the Government of Kenya and the National for sharing samples of the bison skull with us; Tom Guilderson of
Council for Science and Technology (NCST/RRI/12/1/BS- the Lawrence Livermore National Laboratory for assistance with
011/25) in affiliation with the National Museums of Kenya. radiocarbon dating; and the two reviewers who helped greatly
improve this manuscript.
AUTHOR CONTRIBUTIONS
SUPPLEMENTARY MATERIAL
All authors conceived of this study. AS, CD, ER, and DP collected
the field data. LP collected the algal data. AS analyzed the data The Supplementary Material for this article can be found online
and wrote the initial draft of the manuscript. All authors assisted at: https://www.frontiersin.org/articles/10.3389/fevo.2020.00031/
with writing and approving the final manuscript. full#supplementary-material
REFERENCES Bump, J. K., Webster, C. R., Vucetich, J. A., Peterson, R. O., Shields, J. M.,
and Powers, M. D. (2009). Ungulate carcasses perforate ecological filters and
Ameca y Juárez, E. I., Mace, G. M., Cowlishaw, G., and Pettorelli, N. (2012). Natural create biogeochemical hotspots in forest herbaceous layers allowing trees a
population die-offs: causes and consequences for terrestrial mammals. Trends competitive advantage. Ecosystems 12, 996–1007. doi: 10.1007/s10021-009-
Ecol. Evol. 27, 272–277. doi: 10.1016/j.tree.2011.11.005 9274-0
Andrews, P. (1995). Experiments in taphonomy. J. Archaeol. Sci. 22, 147–153. Bunn, S. E., Leigh, C., and Jardine, T. D. (2013). Diet-tissue fractionation of
doi: 10.1006/jasc.1995.0016 δ15N by consumers from streams and rivers. Limnol. Oceanogr. 58, 765–773.
APHA (2006). Standard Methods for the Examination of Water and Wastewater. doi: 10.4319/lo.2013.58.3.0765
Washington, DC: American Public Health Association. Chaloner, D. T., Wipfli, M. S., and Caouette, J. P. (2002). Mass loss and
Atkinson, C. L., Capps, K. A., Rugenski, A. T., and Vanni, M. J. (2016). Consumer- macroinvertebrate colonisation of Pacific salmon carcasses in south-eastern
driven nutrient dynamics in freshwater ecosystems: from individuals to Alaskan streams. Freshw. Biol. 47, 263–273. doi: 10.1046/j.1365-2427.2002.
ecosystems. Biol. Rev. 92, 2003–2023. doi: 10.1111/brv.12318 00804.x
Baco, A. R., and Smith, C. R. (2003). High species richness in deep-sea Christensen, A. M., and Myers, S. W. (2011). Macroscopic observations of the
chemoautotrophic whale skeleton communities. Mar. Ecol. Prog. Ser. 260, effects of varying fresh water pH on bone. J. Forensic Sci. 56, 475–479. doi:
109–114. doi: 10.3354/meps260109 10.1111/j.1556-4029.2010.01646.x
Bar-On, Y. M., Phillips, R., and Milo, R. (2018). The biomass distribution on Earth. Cornwell, W., Weedon, J., and Guofang, L. (2014). litterfitter: Fit a Collection of
Proc. Natl. Acad. Sci. U.S.A. 115, 6505–6511. doi: 10.1073/pnas.1711842115 Models to Single Cohort Decomposition Data. Available at: https://github.com/
Bauer, S., and Hoye, B. J. (2014). Migratory animals couple biodiversity and cornwell-lab-unsw/litterfitter (accessed March 30, 2019).
ecosystem functioning worldwide. Science 344:1242552. doi: 10.1126/science. Cornwell, W. K., and Weedon, J. T. (2014). Decomposition trajectories of diverse
1242552 litter types: a model selection analysis. Methods Ecol. Evol. 5, 173–182. doi:
Beasley, J. C., Olson, Z. H., and Devault, T. L. (2012). Carrion cycling in food webs: 10.1111/2041-210X.12138
comparisons among terrestrial and marine ecosystems. Oikos 121, 1021–1026. Currey, J. D. (2002). Bones: Structure and Mechanics. Princeton, NJ: Princeton
doi: 10.1111/j.1600-0706.2012.20353.x University Press.
Behrensmeyer, A. K. (1978). Taphonomic and ecologic information from bone Edgreen, R. A., Edgren, M. K., and Tiffany, L. H. (1953). Some North American
weathering. Paleobiology 4, 150–162. doi: 10.1017/s0094837300005820 turtles and their epizoophytic algae. Ecology 34, 733–740. doi: 10.2307/193
Behrensmeyer, A. K. (1982). Time resolution in fluvial vertebrate assemblages. 1336
Paleobiology 8, 211–227. doi: 10.1017/s0094837300006941 Ellis, P. S., Shabani, A. M. H., Gentle, B. S., and McKelvie, I. D. (2011). Field
Behrensmeyer, A. K. (1988). Vertebrate preservation in fluvial channels. measurement of nitrate in marine and estuarine waters with a flow analysis
Palaeogeogr. Palaeoclimatol. Palaeoecol. 63, 183–199. doi: 10.1016/0031- system utilizing on-line zinc reduction. Talanta 84, 98–103. doi: 10.1016/j.
0182(88)90096-X talanta.2010.12.028
Behrensmeyer, A. K. (2007). “Bonebeds through time,” in Bonebeds: Genesis, Elser, J. J., Dobberfuhl, D. R., MacKay, N. A., and Schampel, J. H. (1996). Organism
Analysis, and Paleobiological Significance, eds R. R. Rogers, D. A. Eberth, and size, life history, and N:P stoichiometry. Bioscience 46, 674–684. doi: 10.2307/
A. R. Fiorillo, (Chicago, IL: The University of Chicago Press). 1312897
Bergey, E. A., and Cooper, J. T. (2015). Shifting effects of rock roughness across a Fey, S. B., Siepielski, A. M., Nusslé, S., Cervantes-Yoshida, K., Hwan, J. L., Huber,
benthic food web. Hydrobiologia 760, 69–79. doi: 10.1007/s10750-015-2303-4 E. R., et al. (2015). Recent shifts in the occurrence, cause, and magnitude
Frontiers in Ecology and Evolution | www.frontiersin.org 15 February 2020 | Volume 8 | Article 31You can also read
